This guide covers how to use Ouroboros with the Hermes Agent as an execution runtime.
To use Hermes with Ouroboros, ensure you have the Hermes CLI installed (v0.8.0 or higher):
# Verify installation
hermes versionRun the Ouroboros setup command and select the hermes runtime:
ouroboros setup --runtime hermesThis will:
- Configure
~/.ouroboros/config.yamlto use thehermesbackend. - Install Ouroboros skills into
~/.hermes/skills/autonomous-ai-agents/ouroboros/. - Register the Ouroboros MCP server in
~/.hermes/config.yaml.
Once configured, Ouroboros will use Hermes as the orchestrator runtime backend. This does not rewrite llm.backend; interview, ambiguity scoring, and other LLM-only flows continue to use the configured LLM adapter.
ouroboros run seed.yaml --runtime hermesYou can use the ooo command prefix inside a Hermes session to trigger Ouroboros skills:
hermes chat -q "ooo interview 'Build a new CLI tool'"
hermes chat -q "ooo run seed.yaml"Hermes uses the shared stateless ouroboros.router resolver for exact ooo
and /ouroboros: skill dispatch. Adding or changing a command only requires
updating the relevant SKILL.md frontmatter; the runtime keeps logging,
message assembly, and MCP invocation local. See
Shared ooo Skill Dispatch Router.
You can customize the Hermes CLI path in ~/.ouroboros/config.yaml:
orchestrator:
runtime_backend: hermes
hermes_cli_path: ~/.local/bin/hermesOuroboros tracks Hermes sessions using the session_id emitted by the Hermes CLI in quiet mode (-Q). This allows Ouroboros to resume conversations using the --resume flag.
Ouroboros parses the Hermes CLI output to extract the final response and session metadata. It automatically strips reasoning blocks and banners when running in programmatic mode.